UnitedGerbilNation's drive to the Top

[Ragnar Dan]

I see 4393ppd on the 9600GSO, so 3600 does not seem too far off? Where's the 7000ppd? Anyway, may be your GT240 is chewing on a slow WU?

The 7000 is just the figure I've seen for one of those 3470 WU's on the 9600GSO. At least that's the peak. It takes another second per frame every once in a while. In the screenshot, my 8800GT is running a 3470 at over 8000 PPD, while the GTX 260 is getting only 6865 there, though when I use the machine for browsing and posting it slows its progress a little bit.

Maybe Gerbil Jedidiah had an EARLY_UNIT_END he hasn't noticed? That'll slow it down for sure.

Edit: My 9600GSO's settings are: Core = 600, memory = 900 (* 2 = 1800 @DDR3), shader: stock 1700, now at 1735.

[Flying Fox]

My 785G system seems to be drawing more, but I forgot if I got those numbers when I had a HD4550 plugged in. I will measure again once I get my Kill-A-Watt back from a friend who I lent it to.

Idle with 785G and X2 250 @3.0GHz (stock and CnQ'ed) and undervolted to 1.280V: ~50W
Folding with VMware server and Ubuntu 9.10 at low priority: ~90W
The CPU can go up to 3.6GHz but it needs 1.36V and power consumption shoots up by 30W+ so I am not going there.

It is doing those 60xx's at about 22-23 hours with total points (including bonus) of around 1630-1760 each. According to HFM.net I'm doing around ~1800ppd for most projects, vs the E6300+G33 platform which is closer to 1650ppd. They are really close and I bet if I overclock the E6300 to the same GHz they will perform the same. As I said no more SSE discrimination against AMD CPUs for these A3's, which is a good thing.

GJ, do you have numbers when you use the 780G IGP? That way we can deduce the delta that the discrete card adds and if the additional power is worth the points.

UGN is a collective, therefore the cost is distributed among its members. Is that not the purpose of forming this collective?

That's my point! If you can rally a lot of people with idle computers to fold, that is a much more efficient use of power than custom building folding machines that otherwise have no purpose.

There are many issues, the unfortunate thing is that it ended up just a few heavy hitters contributing to the collective and when one has a down period the results suffer significantly. For the past few years it has been especially problematic with the crappy SMP and still-not-up-to-snuff GPU clients that even Damage said he had a hard time making a recommendation to the larger gerbil audience. Let's hope that the new SMP client once it is out of beta will bring enough improvements for us to do more effective recruitment again.

[thecoldanddarkone]

It looks like I might be folding a bit again. I've got an 8800gt that my brother gave me. My Athlon x3+mobo are in the mail.

[Gerbil Jedidiah]

It looks like I might be folding a bit again. I've got an 8800gt that my brother gave me. My Athlon x3+mobo are in the mail.

Excellent! That hardware should be good for 6k to 8k PPD. Maybe even more if you can meet the deadlines on the a3 cores and get the bonuses.

Are you planning on unlocking the 4th core and overclocking?

[Ragnar Dan]

It looks like I might be folding a bit again. I've got an 8800gt that my brother gave me. My Athlon x3+mobo are in the mail.

Good for you. But watch out for the 5910 WU's. Lately they seem to be dominating everything, and they are lower PPD compared to several others.

[thecoldanddarkone]

It looks like I might be folding a bit again. I've got an 8800gt that my brother gave me. My Athlon x3+mobo are in the mail.

Excellent! That hardware should be good for 6k to 8k PPD. Maybe even more if you can meet the deadlines on the a3 cores and get the bonuses.

Are you planning on unlocking the 4th core and overclocking?

We'll see when it gets here, I purchased a fairly cheap mobo. I'll see how much headroom (if any) it has. I'm going to aim for 4-5k. I figure with using the machine 3-4 hours a day that's a pretty decent target.

[Gerbil Jedidiah]

GJ, do you have numbers when you use the 780G IGP? That way we can deduce the delta that the discrete card adds and if the additional power is worth the points.

I'll take care of this when I move over the GT 240 to my main rig while the 285 is getting RMA'd.

Just considering the wattage delta from idle to load, this card gets about 100 PPD per Watt. What are other people's cards doing?

This is important as I have several PCI-E slots I could throw a few cards in. Gimme data everyone!

As exmple, I can get 3 GT 240s for $240. They would be worth 11,000 to 12,000 PPD and pull around 120 Watts.

Really I think I need to swap out the 780G board for something with more PCI-E slots. That would give me 2 or 3 more cards to fold with.


This guy here http://www.newegg.com/Product/Product.aspx?Item=N82E16813186149 could hold 4 GT240s

[Gerbil Jedidiah]

Guys.... REALLY? are you really gonna let the PPD get this low? I don't want to eat a Gerbil.



I'm only folding with 3 PCs for jimminy crimminy's sakes and all...

[flybywire]

A little reprieve for you gerbils as I will be losing two of my bigadv clients in a few days, as the hosting server is going into production. Hopefully I'll be able to move one of those clients onto another server sometime next week. My ppd will definitely take a big hit until then.

[Ragnar Dan]

They're under no threat from me until I get my 9600GSO working again under Linux/Wine. The guides I've read to such things are not updated for changes in software, and they're broken even if you try to use all the old drivers and CUDA. Even if I do get it working, they say it's noticeably slower than under Windows, and I also won't have the capacity to overclock it unless I run Gnome or something (I forget which one it is, actually, since I'm not at that point yet). On the upside, at least I've got the SSH server running successfully on it so I don't have to break my neck trying to read a monitor and use a keyboard connected to the machine, and Samba is working OK, too, so I can monitor the SMP client.

Edit: If I can't get this GPU to run under Wine, I wonder if Win2000 will work. It's the only other OS I have available, and it is not too different fom XP, with the exception of DirectX backwardness as far as I know for the purposes of GPU folding (even though I understand GPU folding doesn't actually call DX functions). And oh yeah, I think I'd have to make a smaller parition for it, too.

[Gerbil Jedidiah]

They're under no threat from me...

Check your SIX!

[Ragnar Dan]

Yeah I've seen that for 2-3 days, now. Win2K looks like a dead end as I figured it would be, from searching around.

[Ragnar Dan]

Maaaaaaybe I got the 9600GSO running now. It didn't crash right away with the UNSTABLE_MACHINE error this time. I think the first time I had the 64 bit CUDA toolkit installed. Now I have to screw around with nice because it hasn't made any progress after 6 minutes on a WU that should take under 2 minutes, though that was 2% overclocked speed .

[Gerbil Jedidiah]

Maaaaaaybe I got the 9600GSO running now. It didn't crash right away with the UNSTABLE_MACHINE error this time. I think the first time I had the 64 bit CUDA toolkit installed. Now I have to screw around with nice because it hasn't made any progress after 6 minutes on a WU that should take under 2 minutes, though that was 2% overclocked speed .

In case it's hardware related Newegg has open box 9600GSOs and GT240s right now. The 9600GSO is $50 with shipping, not too shabby. The GT240s are around 80, so there aren't really a great deal.

I'm loading their video card open box page probably 100 times a day trying to snag a 5850 for around $220... It's kinda like fishing.

[Ragnar Dan]

It's running, about 6:48 a frame instead of 1:42 on the ubiquitous Project 5910. An 80% speed reduction is not pretty.

As for checking 100 times a day, that's 12.5 per hour in an 8 hour day, and that's... ridiculous.

Edit: Now, when I'm logged off the system entirely and leave it running... it's doing over 9:30 per frame.

I think I need a Windows DVD, yeesh.

[Ragnar Dan]

So uh, if I wanted to, could I use the Win7 DVD to upgrade XP with? I think I read about that issue somewhere on here.

(And for the time being I'm killing the SMP client until I can get "nice" to work for me, if that's possible. Maybe in one of the gazillion updates for Ubuntu they'll get around to it eventually. For the time being I can't run Wine + the GPU client using nice at all. It crashes with a permission error, so it's just at whatever priority it defaults to. I ran the SMP client at the lowest priority I could attempt to set it to, which is 19, and it still ate up the 9600GSO's production. With it shut down, it runs a P 5910 at 1:47 or so per frame. Slower, but much closer to Windows speeds.)

(Lots of ninja edits.)

[fuzzhead]

Ragnar Dan

This is the link I used to set up folding in Ubuntu 9.04 Linux on SMP an GPU.

http://www.overclockers.com/forums/show ... p?t=619931

Worked for me!

Fuzzhead

[Ragnar Dan]

Thanks, fuzzhead. 9.04 seems to have fixed my problems. I may have to reinstall it because I used the Server version like the foldingforum thread that I originally used advises, and the link you provided assumes Desktop version without saying so directly in the opening post, which naturally installs a UI, which will apparently let me edit an X11 config file to allow overclocking the GPU, which may help a bit, though I suspect it only allows core overclocking and not shader. It seem slightly slower on GPU folding than my Windows version, but it's close enough that it may just be the slight (just over 2%) difference in shader speed, and the SMP folding is faster, so it works out pretty well.

Very helpful.

[fuzzhead]

Glad you were able to get it running. You are right the GPU is slightly slower
but the SMP is faster so it balances out. Don't try with SLI or you really will have to
edit a bunch of X11. And Ubuntu has better ROI than Win 7, 0 $ versus $$$.

Fuzzhead

[Gerbil Jedidiah]

Well, my PPD is about to go down 8k to 9k. I just boxed up my GTX285 and will be shipping it out tomorrow for RMA.

I'm currently running an ATI3450. The HORROR!!!

[Flying Fox]

Well, my PPD is about to go down 8k to 9k. I just boxed up my GTX285 and will be shipping it out tomorrow for RMA.

I'm currently running an ATI3450. The HORROR!!!

Ouch, the 3450 is even worse than folding single core CPU client. May as well just leave it alone.

[Gerbil Jedidiah]

Well, my PPD is about to go down 8k to 9k. I just boxed up my GTX285 and will be shipping it out tomorrow for RMA.

I'm currently running an ATI3450. The HORROR!!!

Ouch, the 3450 is even worse than folding single core CPU client. May as well just leave it alone.

Yeah I'm not even trying to fold with it. It's good for web-surfing though=)

To help make up for the reduced PPD while my GTX285 is out I have installed the VMWare 3.0 / UNIX combo. My first WU is a BIGADV one. It looks like it will take me 2.37 (34 min, 10 sec per step) days to fold it and will give me 24k PPD. i7 920 running at 3.8GHZ. I mention this to see if this is what others are getting, and if not, maybe there is something I need to tweak?

[Ragnar Dan]

It looks like you had a bunch of WU's for small number of points at the 3 PM update yesterday on EOC, which often means a crashing GPU or some such thing.

[Gerbil Jedidiah]

It looks like you had a bunch of WU's for small number of points at the 3 PM update yesterday on EOC, which often means a crashing GPU or some such thing.

I had not noticed that. I checked my two GPUs. They seem to be working just fine. I was messing around a little bit yesterday with VMware and SMP clients on one rig. Maybe that caused some issues.

[Ragnar Dan]

OK, well, I'm not seeing any large -bigadv points showing up in recent days, so it looks like you've still got something wrong. Are you running with the passkey set on your SMP clients? It looks like all your recent updates are for small point values.

[Gerbil Jedidiah]

OK, well, I'm not seeing any large -bigadv points showing up in recent days, so it looks like you've still got something wrong. Are you running with the passkey set on your SMP clients? It looks like all your recent updates are for small point values.

You may be right. I just looked and my Athlon X2 system was choking on an a3 WU. It had been working on it a day and had 5 days to go. I just killed it. I turned off scientific methods hoping I'd get an a1, but no, I got an a3 again... I just killed the SMP client on that one. I may try it again later to see if some a1 cores are available.

As far as my i7 rig, the first BigAdv WU is at 97%. I've been working on it for 2 days now. It should hit before end of day at least. I'm curious if the a2 or a3 core offers better PPD overall. Any theories?

After all is said and done I have two GPU2 clients and teo SMP2 clients running. I bought a GT 240 last night. Should be here Wednesday, and then I'll be folding with that as well.

[Flying Fox]

OK, well, I'm not seeing any large -bigadv points showing up in recent days, so it looks like you've still got something wrong. Are you running with the passkey set on your SMP clients? It looks like all your recent updates are for small point values.

You may be right. I just looked and my Athlon X2 system was choking on an a3 WU. It had been working on it a day and had 5 days to go. I just killed it. I turned off scientific methods hoping I'd get an a1, but no, I got an a3 again... I just killed the SMP client on that one. I may try it again later to see if some a1 cores are available.

As far as my i7 rig, the first BigAdv WU is at 97%. I've been working on it for 2 days now. It should hit before end of day at least. I'm curious if the a2 or a3 core offers better PPD overall. Any theories?

After all is said and done I have two GPU2 clients and teo SMP2 clients running. I bought a GT 240 last night. Should be here Wednesday, and then I'll be folding with that as well.

For my dual cores, the A2's (not bigadv) are slightly less than the A3's ppd wise, provided I can finish the WU at the fastest time possible (minimal load on the computer). The difference is like 200-300ppd. I suppose for quads the difference should be more pronounced.

[Gerbil Jedidiah]

For my dual cores, the A2's (not bigadv) are slightly less than the A3's ppd wise, provided I can finish the WU at the fastest time possible (minimal load on the computer). The difference is like 200-300ppd. I suppose for quads the difference should be more pronounced.

It doens't really seem to do much for my Q9450.

I'm a fan of those bigadv WUs now though. http://folding.extremeoverclocking.com/ ... =&u=500232

[Gerbil Jedidiah]

Woohoo! GT 240 came in today. No more 3450 for me. It plays Call of Duty, WaW pretty well and FahMon shows it getting 4800 PPD right now on a 5770 WU. All in all a very nice card for the price (69.99 after MIR).

If you guys are thinking of adding a video card for folding I think the GT 240 is easy to recommend. Very nice power to PPD ratio.

[Flying Fox]

Woohoo! GT 240 came in today. No more 3450 for me. It plays Call of Duty, WaW pretty well and FahMon shows it getting 4800 PPD right now on a 5770 WU. All in all a very nice card for the price (69.99 after MIR).

If you guys are thinking of adding a video card for folding I think the GT 240 is easy to recommend. Very nice power to PPD ratio.

So how much more power am I looking at over using an IGP?